Boundary Conditions

Boundary conditions are part of the definition of a wave-mechanics problem. A Hamiltonian written as a differential expression does not by itself determine the spectrum. The allowed wavefunctions must also satisfy endpoint, regularity, matching, periodicity, or decay conditions appropriate to the physical system.

For example, the expression

$$-\frac{\hbar^2}{2m}\frac{d^2}{dx^2}$$

can describe a free particle on the line, a particle in a box, a particle on a ring, or a half-line problem. The difference lies in the Hilbert space and boundary conditions.

Required background. Coordinate Representation supplies the differential-operator language. Continuity Equation supplies the normal-flux and regional norm checks.

Helpful background. Self-Adjoint Operators places the boundary form in its general domain-theoretic setting.

Why Boundary Conditions Matter

Boundary conditions do three jobs:

• they encode physical constraints such as walls, rings, interfaces, and regularity;
• they determine which solutions of the differential equation are allowed;
• they help make the Hamiltonian a valid observable with real energies and unitary time evolution.

In bound-state problems, boundary conditions often quantize energy. In scattering problems, matching conditions determine reflection and transmission amplitudes. In singular potentials, boundary conditions can contain the physical effect of an idealized short-range interaction.

The Boundary Form

For the kinetic operator on an interval $[a,b]$,

$$\hat T = -\frac{\hbar^2}{2m}\frac{d^2}{dx^2},$$

integration by parts gives

$$\begin{aligned} \langle\phi|\hat T\psi\rangle -\langle\hat T\phi|\psi\rangle = -\frac{\hbar^2}{2m} \left[ \phi^*\psi' -\phi'^*\psi \right]_{a}^{b}. \end{aligned}$$

The bracket is the boundary form. A self-adjoint domain must make it vanish for every pair $\phi,\psi$ in that domain and must be maximal with that property. This is the common structure behind familiar boundary conditions.

For one wavefunction, the same combination is proportional to probability current:

$$j = \frac{\hbar}{2mi} \left(\psi^*\psi'-\psi'^*\psi\right).$$

Boundary-form cancellation and probability-flux balance are therefore two views of the same requirement.

Standard Endpoint Families

At an endpoint, the most familiar separated conditions are:

Type	Condition	Typical interpretation
Dirichlet	$\psi=0$	ideal hard wall
Neumann	$\psi'=0$	reflecting endpoint with zero slope
Robin	$\psi'+\gamma\psi=0$	mixed reflecting boundary with real $\gamma$

For real $\gamma$, each Robin endpoint has zero local current because $\psi'/\psi$ is real where $\psi\ne0$. Dirichlet and Neumann are limiting special cases.

Endpoints can also be coupled rather than treated separately. Periodic, twisted-periodic, and more general self-adjoint conditions relate data at $a$ and $b$. Not every arbitrary pair of linear endpoint equations defines a self-adjoint Hamiltonian.

Infinite Walls

An infinite wall excludes the particle from a region. For the infinite square well on $0\lt x\lt L$, the wavefunction must vanish at the walls:

$$\psi(0)=0, \qquad \psi(L)=0.$$

These are Dirichlet boundary conditions. They are not imposed because all walls make wavefunctions vanish. They are imposed because the potential is idealized as infinite outside the interval. For a finite wall, the wavefunction usually extends into the classically forbidden region and decays.

The derivative need not vanish at a hard wall. Indeed, the nontrivial box eigenfunctions have nonzero slope at one or both endpoints. It is also misleading to derive the hard-wall condition by applying the finite-step matching rules and then simply sending the step height to infinity: the domain changes in that limit. The interval problem with Dirichlet endpoints is the clean definition of the idealized infinite well.

Finite Potential Jumps

For a finite step or finite barrier, the wavefunction is continuous across the interface:

$$\psi(x_0^-)=\psi(x_0^+).$$

If the potential is finite at the jump, the derivative is also continuous:

$$\psi'(x_0^-)=\psi'(x_0^+).$$

These conditions follow by integrating the Schrödinger equation over a small interval around $x_0$. A finite jump in $V(x)$ does not produce an infinite impulse in the derivative.

For constant mass, integration of the stationary equation gives

$$-\frac{\hbar^2}{2m} \left[ \psi'(x_0+\epsilon)-\psi'(x_0-\epsilon) \right] + \int_{x_0-\epsilon}^{x_0+\epsilon} (V-E)\psi\,dx =0.$$

If $V$ and $\psi$ remain bounded, the integral vanishes as $\epsilon\to0$, so $\psi'$ is continuous. A jump in $\psi$ itself would generate distributional terms in $\psi''$ that have nothing to cancel them, so $\psi$ is continuous as well.

This rule has assumptions. It applies to the constant-mass Schrödinger operator with no singular interaction at the interface. In effective-mass models, an operator such as

$$-\frac{\hbar^2}{2} \frac{d}{dx} \left( \frac{1}{m(x)}\frac{d}{dx} \right)$$

instead leads, for the simplest interface model, to continuity of $m^{-1}\psi'$ rather than $\psi'$ itself. The microscopic model and operator ordering must be stated before borrowing a matching rule.

Delta-Function Potentials

For a real delta coupling $\lambda$, the potential

$$V(x)=\lambda\delta(x),$$

the wavefunction remains continuous at the origin:

$$\psi(0^-)=\psi(0^+)=\psi(0).$$

The derivative has a jump:

$$\psi'(0^+)-\psi'(0^-) =\frac{2m\lambda}{\hbar^2}\psi(0).$$

This jump condition is the physical content of the singular potential. Requiring derivative continuity here would erase the delta interaction.

Periodic Boundary Conditions

A particle on a ring of circumference $L$ satisfies periodic boundary conditions:

$$\psi(x+L)=\psi(x).$$

For the kinetic-energy Hamiltonian, one also matches derivatives:

$$\psi'(x+L)=\psi'(x).$$

Periodic boundary conditions encode topology: $x=0$ and $x=L$ are the same physical point. They produce momentum quantization of the form

$$k_n=\frac{2\pi n}{L}, \qquad n\in\mathbb Z.$$

The plane-wave basis, box normalization, and density-of-states limit are later applications of this endpoint identification.

A broader self-adjoint family uses a common phase at the identified endpoints:

$$\begin{aligned} \psi(L)&=e^{i\theta}\psi(0),\\ \psi'(L)&=e^{i\theta}\psi'(0). \end{aligned}$$

These are twisted-periodic conditions. For a plane wave $e^{ikx}$ they give

$$k_n=\frac{2\pi n+\theta}{L}, \qquad n\in\mathbb Z,$$

where only $\theta$ modulo $2\pi$ matters. Ordinary periodicity is the case $\theta=0$. Twists also arise when a gauge potential is traded for a boundary phase, but the Hamiltonian and gauge convention must then be transformed consistently.

Worked Example: A Robin Wall

Consider a free particle on $0\le x\le L$ with

$$\psi(0)=0, \qquad \psi'(L)+\gamma\psi(L)=0,$$

where real $\gamma$ has dimensions of inverse length. For a positive energy $E=\hbar^2k^2/(2m)$, the left condition selects

$$\psi(x)=A\sin(kx).$$

The Robin condition then gives the spectral equation

$$k\cos(kL)+\gamma\sin(kL)=0.$$

Thus the boundary parameter changes the allowed wave numbers even though the differential equation in the interior is unchanged. At $\gamma=0$, the right endpoint is Neumann and $\cos(kL)=0$. For large positive $\gamma$, the roots approach the Dirichlet values $\sin(kL)=0$.

Boundary conditions can do more than shift positive-energy levels. If $\gamma\lt-1/L$, this example also has one negative-energy solution. Writing $E=-\hbar^2\kappa^2/(2m)$ gives

$$\gamma=-\kappa\coth(\kappa L),$$

which describes a state localized near the Robin endpoint. This is a useful warning that endpoint physics can act like an interaction.

Decay And Regularity

Bound states on the full line must be square integrable:

$$\int_{-\infty}^{\infty}\lvert\psi(x)\rvert^2\,dx\lt\infty.$$

This usually requires decay at spatial infinity. A candidate energy is a bound-state eigenvalue only if the corresponding solution lies in the Hilbert-space domain of the Hamiltonian; merely finding an exponentially decaying branch on one side is not enough.

In radial problems, the origin is an endpoint of a half-line problem rather than an ordinary interior point. If $u_\ell(r)=rR_\ell(r)$ is the reduced radial wavefunction, the standard nonsingular three-dimensional problem normally requires $u_\ell(0)=0$. Singular central potentials can require a more careful domain analysis, so one should not infer the origin condition from square integrability alone. The full radial classification remains a later model-specific treatment.

Decay and regularity conditions are boundary conditions too, even when no hard wall is visible.

Interface Conditions Versus Scattering Conditions

Matching at a finite interface and selecting a scattering experiment are logically different steps. Suppose a potential is localized near the origin. A left-incident stationary solution is usually written asymptotically as

$$\psi(x)\sim \begin{cases} e^{ikx}+r e^{-ikx}, & x\to-\infty,\\ t e^{ikx}, & x\to+\infty. \end{cases}$$

Continuity conditions at the interfaces follow from the local Schrödinger equation. The absence of an incoming wave from the right instead selects which generalized eigenfunction represents the experiment. It is not a new endpoint domain for the self-adjoint full-line Hamiltonian.

The distinction matters for resonances. Imposing purely outgoing behavior on both sides and analytically continuing the energy can produce complex resonance poles. Those solutions are valuable descriptions of decay, but they are not normalizable eigenstates with complex eigenvalues of the original closed, self-adjoint Hamiltonian.

Boundary Conditions And Current

Acceptable boundary conditions should be compatible with probability conservation. In one dimension, the probability current is

$$j =\frac{\hbar}{2mi} \left(\psi^*\frac{d\psi}{dx} -\psi\frac{d\psi^*}{dx}\right).$$

For an isolated box with hard walls, no probability should flow through the walls. For a ring, current can circulate, but the wavefunction must match consistently around the loop. For scattering states, incoming, reflected, and transmitted currents determine physical probabilities.

The continuity equation makes the endpoint statement precise:

$$\frac{\partial\rho}{\partial t} +\frac{\partial j}{\partial x}=0, \qquad \frac{d}{dt} \int_a^b \rho\,dx =j(a)-j(b).$$

Dirichlet, Neumann, and real Robin conditions make $j$ vanish separately at each isolated endpoint. Periodic and twisted-periodic conditions instead give $j(a)=j(b)$: probability may circulate through the identified endpoint, but none is lost. Coupled boundary conditions therefore need not set each endpoint current to zero; they must balance the total flux.

This current viewpoint is a practical way to detect boundary conditions that would not describe a closed quantum system.

Boundary Conditions And Self-Adjointness

At a deeper level, boundary conditions define the domain of the Hamiltonian. A differential expression can be symmetric on a chosen domain without yet being self-adjoint. Self-adjointness requires the domain of the operator to equal the domain of its adjoint. It is this stronger property that supports a real spectral resolution and unitary evolution generated by the Hamiltonian.

The boundary form is the practical diagnostic in regular one-dimensional examples. A proposed domain must make that form vanish for every pair of states in the domain, and the conditions must be complete rather than unnecessarily restrictive. For introductory wave mechanics, the durable rule is:

Do not specify only the differential expression. Specify the allowed wavefunctions.

Self-Adjoint Operators develops the rigorous operator-domain criterion used here as a check.

Numerical Implementation

A discretization should preserve the same domain logic as the continuum problem. Common implementations include:

• removing fixed Dirichlet endpoint values from the vector of unknowns;
• connecting the first and last grid points with wrap-around matrix elements for periodic conditions;
• eliminating ghost points for Neumann or Robin conditions in a way that preserves the discrete inner product and matrix symmetry;
• choosing basis functions that satisfy the desired conditions before diagonalization.

An arbitrary one-sided endpoint stencil can make an otherwise real finite-difference Hamiltonian non-Hermitian. Before trusting its spectrum or time evolution, check

$$\lVert H-H^\dagger\rVert,$$

verify that low-lying eigenvalues converge as the grid is refined, and confirm the expected endpoint or interface current balance. On nonuniform grids or in finite-element methods, Hermiticity must be assessed with the appropriate quadrature or mass matrix rather than the unweighted Euclidean dot product.

Common Mistakes

• Imposing $\psi=0$ at every potential jump.
• Requiring $\psi'=0$ at an infinite wall.
• Requiring $\psi'$ to be continuous across a delta-function potential.
• Applying constant-mass derivative matching to a position-dependent-mass Hamiltonian without deriving the interface rule.
• Forgetting derivative matching for periodic kinetic-energy problems.
• Giving $\psi$ and $\psi'$ different endpoint phases in a twisted-periodic problem.
• Solving a bound-state equation without checking square integrability.
• Treating incoming or outgoing scattering conventions as though they changed the self-adjoint domain of the closed Hamiltonian.
• Treating boundary conditions as optional after finding a general solution.
• Using the same boundary conditions for finite and infinite barriers.
• Discretizing an endpoint condition without checking the resulting matrix for Hermiticity.

Where this is used

• Time-Independent Schrödinger Equation becomes a spectral problem only after boundary conditions are stated.
• Hamiltonians in Coordinate Space separates the differential expression from its domain.
• Probability Current supplies the flux diagnostic.
• Continuity Equation converts that flux into regional norm balance.

Exercises

1. Derive the derivative jump condition for $V(x)=\lambda\delta(x)$ by integrating the stationary Schrödinger equation across $[-\epsilon,\epsilon]$ and taking $\epsilon\to 0$.

Solution

The stationary equation is

$$-\frac{\hbar^2}{2m}\psi''(x)+\lambda\delta(x)\psi(x)=E\psi(x).$$

Integrate across $[-\epsilon,\epsilon]$:

$$-\frac{\hbar^2}{2m} \left[\psi'(\epsilon)-\psi'(-\epsilon)\right] +\lambda\psi(0) =E\int_{-\epsilon}^{\epsilon}\psi(x)\,dx.$$

The right side vanishes as $\epsilon\to 0$ for a finite wavefunction. Therefore

$$\psi'(0^+)-\psi'(0^-) =\frac{2m\lambda}{\hbar^2}\psi(0).$$

2. Why is $\psi(0)=\psi(L)=0$ appropriate for an infinite square well but not for a finite square well?

Solution

In an infinite well, the outside region is excluded by an infinite potential, so the wavefunction must vanish at the boundaries. In a finite well, the particle can have an evanescent tail in the classically forbidden outside region. The wavefunction and derivative are matched across the finite jump instead of forcing the wavefunction to vanish at the wall.

3. Show that the kinetic-energy boundary form vanishes for (a) Dirichlet conditions and (b) twisted-periodic conditions with the same phase $\theta$ for every state in the domain.

Solution

For Dirichlet conditions, $\phi(a)=\phi(b)=\psi(a)=\psi(b)=0$. Every term in

$$\left[ \phi^*\psi'-\phi'^*\psi \right]_a^b$$

therefore vanishes.

For twisted-periodic conditions,

$$\begin{aligned} \phi(b)&=e^{i\theta}\phi(a), & \phi'(b)&=e^{i\theta}\phi'(a),\\ \psi(b)&=e^{i\theta}\psi(a), & \psi'(b)&=e^{i\theta}\psi'(a). \end{aligned}$$

The upper-endpoint expression is

$$\begin{aligned} \phi^*(b)\psi'(b)-\phi'^*(b)\psi(b) &=e^{-i\theta}\phi^*(a)e^{i\theta}\psi'(a)\\ &\quad-e^{-i\theta}\phi'^*(a)e^{i\theta}\psi(a)\\ &=\phi^*(a)\psi'(a)-\phi'^*(a)\psi(a). \end{aligned}$$

The upper and lower contributions cancel, so the boundary form is zero.

4. A free particle on a ring obeys $\psi(L)=e^{i\theta}\psi(0)$ and $\psi'(L)=e^{i\theta}\psi'(0)$. Derive its allowed wave numbers and energies. Explain why $\theta$ and $\theta+2\pi q$, with integer $q$, describe the same spectrum.

Solution

For a momentum eigenfunction $\psi(x)=Ae^{ikx}$, either endpoint condition gives

$$e^{ikL}=e^{i\theta}.$$

Hence

$$k_n=\frac{2\pi n+\theta}{L}, \qquad E_n=\frac{\hbar^2}{2m} \left( \frac{2\pi n+\theta}{L} \right)^2, \qquad n\in\mathbb Z.$$

Replacing $\theta$ by $\theta+2\pi q$ relabels $n$ as $n+q$. The set of energies is therefore unchanged.

5. For the free particle on $[0,L]$ with $\psi(0)=0$ and $\psi'(L)+\gamma\psi(L)=0$, derive the positive-energy quantization condition. Recover the Neumann and Dirichlet limits, and determine when a negative-energy solution exists.

Solution

For $E=\hbar^2k^2/(2m)$, the left endpoint condition reduces the general solution to $\psi=A\sin(kx)$. The right endpoint then requires

$$k\cos(kL)+\gamma\sin(kL)=0.$$

At $\gamma=0$, this becomes $\cos(kL)=0$, the Neumann condition at $L$. As $\gamma\to+\infty$, roots approach $\sin(kL)=0$, the Dirichlet spectrum at $L$.

For $E=-\hbar^2\kappa^2/(2m)$ with $\kappa\gt0$, the left condition gives $\psi=A\sinh(\kappa x)$. The right condition becomes

$$\kappa\cosh(\kappa L) +\gamma\sinh(\kappa L)=0,$$

or

$$\gamma=-\kappa\coth(\kappa L).$$

The function $\kappa\coth(\kappa L)$ increases from $1/L$ to infinity as $\kappa$ increases. A unique negative-energy solution therefore exists precisely when $\gamma\lt-1/L$. At $\gamma=-1/L$, the threshold solution has zero energy but is not a negative-energy state.

References

• G. B. Arfken, H. J. Weber, and F. E. Harris, Mathematical Methods for Physicists, 7th ed., Academic Press, 2013.
• G. Bonneau, J. Faraut, and G. Valent, “Self-adjoint extensions of operators and the teaching of quantum mechanics,” American Journal of Physics 69, 322–331 (2001), doi:10.1119/1.1328351.
• C. Cohen-Tannoudji, B. Diu, and F. Laloë, Quantum Mechanics, Wiley, 1977.
• D. J. Griffiths and D. F. Schroeter, Introduction to Quantum Mechanics, 3rd ed., Cambridge University Press, 2018.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics, Vol. II: Fourier Analysis, Self-Adjointness, Academic Press, 1975.
• R. Shankar, Principles of Quantum Mechanics, 2nd ed., Springer, 1994.